Addressing the Disparity
between Climate Models and
Observations: Testing the
Hypothesis of AGW
Conference on Global and Regional Climate Variability
Santa Fe, NM
Oct 31-Nov 4, 2011
(Prof.) S. Fred Singer
University of Virginia/ SEPP
<[email protected]>
• The reports of the IPCC [2007], of the USClimate Change Science Program [CCSP-SAP1.1 2006], and of the NIPCC [2008] all show a
major disparity between modeled and
observed patterns of temperature trends. In
particular, all GH models show a “hot spot” in
the tropical upper troposphere; however,
radiosonde and independent satellite
observations do not.
Resolving the conflicting claims
• Based on recent papers by Singer [Energy&Envir 2011], by Fu et al
[GRL 2011], and by Thorne et al [JGR 2011], we try to resolve the
current controversy between Douglass et al [DCPS in IJC 2007] and
Santer et al [IJC 2008]. DCPS and Santer report conflicting results
when comparing observed and modeled patterns of temperature
trends ("fingerprints"). For example, DCPS report a disparity, while
Santer claims “consistency” -- that observed temperature trends in
the tropical troposphere agree with trends derived from GeneralCirculation Models (GCMs). Santer’s claim is based mainly on the
visual overlap above 400 hPa, seen in his Fig. 6, between a new set
of tropospheric temperature trends (RAOBCORE, derived from Reanalysis) and the interval of uncertainty in the modeled trend
values, derived from the "20CEN" set of GCMs compiled by the
Intergovernmental Panel on Climate Change (IPCC 2007).
• Like Fu, we find that the "new set of tropospheric temperature
trends" is not in accord with satellite data – contra Santer’s claim.
Further, the "error interval" deduced for the model trends (the grey
area of Santer’s Fig. 6) may simply be an artifact caused by the
presence of many single-run models in the IPCC’s “20CEN”
compilation. If "chaotic uncertainty" of models were eliminated, this
grey envelope would shrink -- and any overlap between the modeled
error interval and observed temperature trends would be less likely.
Thorne et al, which includes Santer as a coauthor, conclude that “…
agreement between models, theory, and observations … is
nonexistent above 300 hPa.”
• Thus, observations and models remain inconsistent -raising questions about IPCC's estimates of "climate
sensitivity" and hence the magnitude of anthropogenic
global warming imagined by the GCMs.
Three Fundamental Issues
• 1. Is climate change of the 20th century
anthropogenic (human-caused)?
• 2. If AGW is significant, is a future warmer
climate (and higher CO2 level) good or bad ?
• 3. If a future warming is calamitous, can we
really do something about it? Can we lower
atmospheric CO2 levels, phase out fossil fuels,
make solar and wind power reliable and
cheap, produce biofuels economically?
Three Major Science Questions
“Gallia est omnis divisa in partes tres”
• 1. Natural or Anthropogenic Forcing?
A Science and also a Policy Issue
• 2. Why the Disparity between Models &
Obs? Negative feedback or Saturation?
• 3. What is Causing Climate Variability?
Internal oscillations or External factors?
Is There Disparity between Modeled
and Observed Temperature Trends?
• Attribution of observed warming trends to
GH-gas increases is based largely on claimed
agreement between the patterns of observed
(tropical) tropospheric trends and modeled
ones [Santer et al., in IPCC 1996, Chap 8, and
in IJC 2008, Fig 6]. We show that the claimed
consistency is spurious.
• The “fingerprints” don’t match.
• 1. IPCC-4 (2007) and CCSP-SAP-1-1 (2006)
• 2. NIPCC Summary (2008)
• 3. Singer (E&E 2011) “Lack of consistency …”
and references therein
From Energy & Environment , vol 22, no 4, pp 375-406, 2011
S Fred Singer
University of Virginia/ SEPP
Email: [email protected]
The US Climate Change Science Program [CCSP, 2006] reported, and
Douglass et al. [2007] and NIPCC [2008] confirmed, a “potentially serious
inconsistency” between modeled and observed trends in tropical surface and
tropospheric temperatures. However, Santer et al. [2008: hereafter “Santer”],
though sharing several co-authors with CCSP [2006], offered “new
observational estimates of [tropical] surface and tropospheric temperature
trends”, concluding that “there is no longer a serious discrepancy between
modelled and observed trends.” Santer’s key graph [shown here as Fig. 5]
misleadingly suggests an overlap between observations and modeled
trends. His “new observational estimates” conflict with satellite data. His
modeled trends are an artifact, merely reflecting chaotic and structural model
uncertainties that had been overlooked. Thus the conclusion of “consistency” is
not supportable and accordingly does not validate model-derived
projections of dangerous anthropogenic global warming (AGW)
CCSP 1.1 – Chapter 1, Figure 1.3F PCM Simulations of
Zonal-Mean Atmospheric Temperature Change
Height (km)
CCSP 1.1 – Chapter 5, Figure 7E
Height (km)
A more detailed view of the disparity:
Douglass, Knox, Pearson, Singer IJC 2007
Reaction to DCPS (IJC 2007)
• Santer et al (IJC 2008) claim “Consistency” of
modeled and observed trends – by:
• 1. Adding a new dataset that has a max trend
in the upper troposphere, at 200 hPa
• 2. Greatly expanding the uncertainty interval
of the modeled trends
But others confirm DCPS(IJC 2007)
and Singer (Energy&Envir 2011)
• Fu, Manabe, Johanson (GRL 2011):
It is "critically important to understand the causes
responsible for the discrepancy between the models
and observations.”
“… trends from [satellites] are significantly smaller
than those from GCMs”
• Thorne, Santer, et al (JGR 2011):
" ... agreement between models, theory, and
observations .. is nonexistent above 300 hPa."
Two different views from CCSP
• A. In Chap 5 (Santer, lead author) shows a
clear disparity between model runs and data.
• B. But the Exec Summary (Wigley, lead author)
suggests overlap (i.e., Consistency), by using
the concept of ‘range’ (i.e., Max--Min of the
distribution) – not a valid statistical metric.
(a) CCSP 1.1 – Chapter 5, Figure 4G
(b) CCSP 1.1 - Executive Summary Figure 4G: Modeled and Observed Temperature Trends in
the Tropics (20oS-20oN)
Model uncertainty (‘grey area’)
• Santer et al (2008) claim that the ‘grey area’ (in
Panel A) denotes 2-sigma uncertainty. Singer
2011 finds that it is the ‘range’ (Max—Min) of
the model runs –an impermissible statistical
metric – thereby greatly inflating the model
• Santer claims (in Panel B) that the new dataset
is supported by satellite data. But Singer (E&E
2011) and Fu, Manabe, Johanson (GRL 2011)
both disagree with Santer.
Trend (K/dec)
1 sims9
2 sims5
3 sims5
4 sims5
5 sims5
6 sims4
7 sims4
8 sims3
9 sims3
10 sims3
11 sims3
12 sims3
13 sims2
14 sims2
15 sims2
16 sims2
17 sims2
18 sims1
19 sims1
20 sims1
21 sims1
22 sims1
Santer trend
Grey left
Grey right
Climate models are chaotic
• E.g., the 5 runs of the Japan MRI model show
trends ranging from 0.042 to 0.371 K/decade.
• Which of these 5 modeled trends should we
compare to the observed trend?
• In a synthetic experiment we show that at
least 40 runs (of 20-yr length) are necessary to
get convergence of the ‘cumulative ensemblemean – and >20 runs of 40-yr long runs.
Cumulative Ensemble-mean trend
vs. Number of runs
• Modeled fingerprints don’t match observed
• Models are not validated; should not be used
as policy tools
• Individual model runs show wide variations of
temp trends – evidence of chaotic behavior
• We need more studies to explain what causes
discrepancy between models and obs
• 1. The US-CCSP report shows major differences between
observed temp trends and those from GH models
• These disagreements are confirmed and extended by
Douglass et al [in IJC 2007] and by NIPCC 2008
• Claims of “consistency’” between models and obs by
Santer et al [in IJC 2008] are shown to be spurious
• 2. IPCC-4 [2007] climate models use an insufficient
number of runs to overcome “chaotic uncertainty”
• 3. We find no evidence in support of the surface warming
trend claimed by IPCC-4 as evidence for AGW
• We conclude that current warming is mostly natural
and that the human contribution is minor.

Addressing the Disparity between Climate Models and Observations